A Spectral Color Imaging System for Estimating Spectral Reflectance of Paint

Size: px
Start display at page:

Download "A Spectral Color Imaging System for Estimating Spectral Reflectance of Paint"

Transcription

1 A Spectral Color Imagng System for Estmatng Spectral Reflectance of Pant Vladmr Bochko Department of Informaton echnology, Lappeenranta Unversty of echnology, P.O.Box 2, Lappeenranta, Fnland Normch sumura and Yoch Myake Department of Informaton and Image Scences, Chba Unversty, 1-33 Yayo-cho, Inage-ku, Chba , Japan ) In ths paper, the analyss methods used for developng magng systems estmatng the spectral reflectance are consdered. he system ncorporates the estmaton technque for the spectral reflectance. Several tradtonal and machne learnng estmaton technques are compared for ths purpose. he accuracy of spectral estmaton wth ths system and each estmaton technque s evaluated and the system s performance s presented. Introducton In ths paper, the analyss methods used for developng magng systems estmatng the spectral reflectance are consdered. he estmaton of the spectral reflectance determnes a performance of a hgh qualty color magng system whch s requred n dgtal archves, network museums, e-commerce and telemedcne. 1 Especally the desgn of a system for accurate dgtal archvng of fne art pantngs has awakened 1

2 ncreasng nterest. In such a system the dgtal mage s easly examned by usng a broadband network. he vstors of museums, art experts and artsts would be able apprecate a varety of pantngs at any vewng ste wherever those pantngs are located. In addton, archvng the current condton of a pantng wth hgh accuracy n dgtal form s mportant to preserve t for the future. Several research groups worldwde have been workng on these problems. 2,3,4,5,6,7,8,9,1,11,12,13,14 Conventonal color magng systems have a lmtaton that s a dependence of mages on the llumnant and characterstcs of the magng system. he magng systems based on spectral reflectance, unlke the conventonal systems, are devcendependent and capable of reproducng the mage of the scene n any llumnaton condtons. Also, these systems can ncorporate the color appearance characterstcs of the human vsual system. Owng to the fact that spectral characterstcs are smoothed, the hgh-dmensonal spectral reflectance s accurately represented by a small number of channel mages. 15,16,17 herefore, the task of spectral estmaton ncludes statstcal analyss of the reflectance spectra and mnmzaton of the estmaton error. he choce of error measures s a general topc of broader nterest and sometmes contrary n mpact. In the archval, ramfcatons for optmzng more for RMSE versus color dfference depend on applcatons. For example, spectral optmzaton may better enable the dentfcaton of colorants used whle color dfference optmzaton may yeld superor vsual reproductons. he tradtonal technques used for the estmaton nvolve matrx-vector computaton and usually assume a lnear model of the data. Although the approach based on lnear algebra and a nonlnear data model s proposed n the lterature, 4 machne learnng 2

3 technques seem appealng. hey estmate spectra of the scene, ncorporate the data nonlnearty and nvolve the tranng and predcton procedures. herefore, the neural networks based methods for spectral reconstructon are proposed by Rbes et al. 18 he tested methods are superor to the pseudo-nverse based estmaton method wth a quantzaton nose. Wthout nose the tradtonal methods predct better than the neural network because of the hghly lnear relatonshp between spectral sets used for tranng and predcton. o provde color constancy a Bayesan approach of the estmaton method s proposed by Branard and Freeman. 19 Snce the Bayesan approach s computatonally demandng, the submanfold method for spectral reflectance estmaton that s an ntermedate soluton between the Bayesan approach and lnear estmaton methods s descrbed by DCarlo and Wandell. 2 he method extends the lnear methods and ntroduces the addtonal term ncorporatng the nonlnearty of the data. he method uses a pece-wse lnear way to represent the nonlnear data structure and reduces the error value 12% n comparson wth a lnear method. It s mportant that the method partcularly reduces large lnear errors. he lmtaton of the method s that t needs a large tranng set and s nsuffcent when the data structure s a one-to-many mappng. he propertes of the methods consdered n ths paper are qute close to the submanfold approach 2 and one of the learnng algorthms based on Wener estmaton also gves a pece-wse lnear soluton. Recently, many advanced machne learnng technques usng neural networks and support vector machnes have been ntroduced and combned n the lbrares that are convenent for the purpose. For example, buldng the estmaton methods usng the ready-made machne learnng algorthms one can get theoretcally founded algorthms, 3

4 a unfed workflow for a current and future study, and a rch set of methods that provde flexblty for applcaton-orented research. In ths paper, the neural networks algorthms from the Netlab lbrary 21, 22 wll be used. hey nclude regresson, clusterng and pattern recognton methods. Many of these methods are densty models based on a lkelhood that s mportant for recognton and convenent for comparson wth other methods. In ths study, we statstcally analyze the reflectance spectra of color-patch sets of ol and watercolor pantngs wthout nose characterstcs, develop three machnelearnng based methods and compare them wth three tradtonal methods wth a synthetc data set and the real color-patch sets, as well. he tradtonal methods are lnear estmators based on low-dmensonal prncpal component analyss (PCA) approxmaton and Wener estmaton, and a nonlnear estmator based on multple regresson approxmaton. he machne learnng methods extend the tradtonal methods for estmatng a nonlnear data structure. hey nclude two nonlnear methods based on nonlnear prncpal component analyss and regresson analyss and the method usng pece-wse lnear Wener estmaton. he method utlzng nonlnear PCA and the method explotng pece-wse lnear Wener estmaton are novel methods. o develop an magng system, two measures are used for estmaton accuracy: spectral color dfference (RMSE) and colormetrc color dfference (CIE E 94 ). he former s better for archvng the spectral reflectance and the latter s better for evaluatng the appearance of the art pantngs under a specfc llumnaton to human observers. 4

5 he paper s arranged as follows: In the followng secton, we formulate the generalzed reconstructon of spectral reflectance from a multchannel mage n magng systems wth a reduced number of channels. Next, we descrbe three tradtonal methods and three machne learnng methods. hen we present the results of the statstcal analyss of the reflectance spectra of the color patches. Later on, an experment wth synthetc data and the reflectance spectra of the color patches s descrbed. Fnally, the expermental results are dscussed and concludng remarks are presented. Formulaton of the Spectral Reflecton Estmaton Fg. 1 shows the mage acquston system. he system conssts of the sngle chp hgh qualty CCD camera and the rotatng color wheel comprsng several color flters. he response v at poston ( x, y) of the CCD camera wth the th color flter s expressed as follows 3 : v ( x, y) = t ( λ ) E( λ) S( λ) r( x, y, λ) dλ + n ( x, y), = 1, K m, (1), where t (λ), E (λ), S (λ) and r ( x, y, λ) are the spectral transmttance of the th flter, the spectral radance of the llumnant, the spectral senstvty of the camera, and the spectral reflectance of a pantng, respectvely. n ( x, y) denotes addtve nose n the th channel mage and m denotes the total number of channels. 5

6 Fg. 1. he mage acquston system. For mathematcal convenence, each spectral characterstc wth l wavelengths s expressed as a vector or a matrx. Usng vector-matrx notaton, we can express Eq. (1) as follows: v ( x, y) = ES r( x, y) + n( x, y), (2) where denotes a transposton, v s an m 1 column vector representng the camera response, r s an l 1 column vector representng the spectral reflectance of the pantng, = t1, t 2, K, t m s an l m matrx n whch each column t represents the transmttance of the th flter, and E, S are the l l matrces that correspond to the spectral radance of the llumnant and the spectral senstvty of the CCD camera, respectvely. Further for the sake of smplcty, ( x, y) from v, r and n are omtted. Eq. (2) s rewrtten as an overall, lnear system matrx F = ES wth m l elements: v = Fr + n. (3) 6

7 he response of the spectral CCD camera v wthout a nose term s as follows: v = Fr. (4) We wll call the space spanned by r a spectral space and the space spanned by v a sensor space or subspace. he estmaton of reflectance spectra s obtaned as follows: r ˆ = Gv, (5) where G s a matrx dependng on the estmaton method used. In the next sectons, sx estmaton methods are consdered. radtonal Estmaton echnques hree approaches are usually used for spectral sensor desgn. he estmaton technques of reflectance spectra nclude: the method based on PCA (low-dmensonal approxmaton) (PCE), the method based on Wener estmaton (WE) and the method usng multple regresson approxmaton (MRE). 4 he Method Based on PCA Usng spectral reflectance of the tranng set r a covarance matrx s computed as follows: where E () s an expectaton operator. C = E(( r E( r))( r E( r)) ), (6) An egendecompston of the covarance matrx C determnes the matrx B = b, b, K, b 1 2 k, the columns of whch are k egenvectors correspondng to the frst k largest egenvalues. he spectral reflectance s approxmated as follows: r Bw, (7) where w s a vector of PCs, w = w 1, w2, K, w k and m k. 7

8 he spectral camera response gven by Eq. 4 can be presented by another expresson as follows 17 : v = FBw. (8) he prncpal components (PCs) are determned as follows: 1 w = ( FB) v. (9) Usng Eq. 7 and Eq. 9 the estmaton matrx G s as follows: 1 G = B( FB). (1) he estmate of the spectral reflectance of the pantng s as follows: 1 rˆ = Gv = B( FB) v, (11) where the data s centered by v v E(Fr) and means that the expresson on the rght s calculated and replaces the expresson on the left. Fnally, the mean value s added as follows: rˆ = rˆ + E( r). (12) Better accuracy of estmaton can be obtaned wth Wener estmaton, whch s consdered next. he Method Usng Wener Estmaton he Wener estmaton method mnmzes the overall average of the square error between the orgnal and estmated spectral reflectance. 3 For ths method, the correlaton matrces R rr of pantng spectra and nose consequently, the estmaton matrx s the followng 3 : R nn are frst computed, and 1 G = RrrF ( FR rrf + Rnn). (13) he estmate s as follows: rˆ 1 Gv = Rrr F ( FR rrf + R ) v. (14) = nn 8

9 If nose s not consdered, the estmaton matrx s as follows 3 : 1 G = R rrf ( FR rrf ). (15) And the estmate s as follows: 1 rˆ = Gv = R rr F ( FR rrf ) v. (16) In ths study, the Wener estmaton wthout consderaton to nose s used. he Wener estmaton gves good accuracy for lnear data. If the data s nonlnear, the technque based on multple regresson analyss s used. he Method Usng Multple Regresson Analyss In the case of nonlnear data, multple regresson analyss gves better results than Wener estmaton. 4 In the MRE method, the extended data matrx V of pantng spectra s frst defned through the data components and ther extended set of hgher-order terms as follows 4 : V = v, K, vm, v v, v v K, hgher order terms,, (17) K where denotes element-wse multplcaton. hen the estmaton matrx s gven as follows: 1 G = RV ( VV ), (18) where R s a matrx, the columns of whch are presented by n spectral samples gven by R = r, r, 2,, (19) 1 K r n Accordng to the lterature 4, the estmaton matrx G used n MRE s equal to the noseless varant of the Wener estmaton matrx. Fnally, 9

10 rˆ = GV = RV ( VV ) 1 V. (2) Owng to the fact that new advanced machne learnng algorthms are especally relevant for workng wth a nonlnear structure of data, the machne learnng technques are next dscussed for spectral estmaton. Machne Learnng Estmaton echnques Drawng analogy wth the tradtonal estmaton methods, three machne learnng technques are proposed. hey nclude the method based on regressve (nonlnear) PCA (RPCE), the method based on pece-wse lnear Wener estmaton (PLWE) and the method usng regresson analyss (RE). Eq.1 Eq.5 are vald for all machne learnng methods. he Method Based on Regressve PCA he spectral camera response s computed n the followng way: v = FBf w, θ ), (21) ( f where f () s a nonlnear vector-valued mappng functon and θ f s a parametrc vector. hen, PCs are defned by the followng equaton 1 w = h(( FB) v, ), (22) θ h where h() s an nverse functon, h () = f () 1, θ h s a parametrc vector and v v E(Fr). he mappng functon h () and parametrc vector θ h are computed usng a machne learnng algorthm for regresson. 21 In consequence, the spectral estmate of the pantng s as follows: 1

11 ˆ 1 r = Bh(( FB) v, ). (23) θ h Fnally, the mean value s added as follows: rˆ rˆ + E( r). (24) In practce, ths method nvolves a low-dmensonal subspace and a hgherdmensonal subspace ncludng the low-dmensonal subspace. For the lowdmensonal subspace, where w ( k ) = w, w, K, w, the mappng s as follows: 1 2 k w ( k ) 1 = h(( FB) v, θ h ) = ( FB) 1 v, (25) where v Fr E(Fr). For the hgher-dmensonal subspace, where ( p) ( k ) ( k + 1: p) w = w, w = w, w,, w, w, K, w, (26) 1 2 K k k + 1 the mappng s done for the hgher-order (or weak) PCs as follows: p w ( k + 1: p) = h(( FB) 1 v, θ h ) = h( w ( k ), θ). (27) hus the method uses the low-order real PCs and the hgher-order approxmated PCs. he Method Usng Pece-Wse Lnear Wener Estmaton In ths secton, the other machne learnng algorthm for pece-wse lnear Wener estmaton s dscussed. he man dea of the method s to separate the data structure nto parts whch are sutable for lnear approxmaton and each part s then estmated by usng the lnear Wener estmaton method. For data separaton, the clusterng algorthm s frst requred. he data s dvded nto several clusters v usng the Gaussan mxture model (GMM) 21 n a sensor space where s an ndex of the cluster. hen for the data of each cluster Wener estmaton s utlzed. Usng the labels of the data t s easy to compute the cluster covarance 11

12 matrx n the spectral doman needed for estmaton. When the th matrx follows: cluster covarance C of pantng spectra s known, the spectral estmate for the th cluster s as rˆ 1 = G v = CF ( FCF ) v, (28) where v v E Fr ). ( Fnally, the mean value s added as follows: r ˆ rˆ + E( r ). (29) he estmaton procedure s sequentally repeated for all clusters. he Method Usng Regresson Analyss he estmaton method based on the regresson analyss s smlar to the multple regresson approach. he dfference s that nonlnear mappng s used nstead of lnear mappng and the hgher-order terms are not syntheszed. For regresson analyss based on machne learnng the estmate s gven as follows: r ˆ = g( v, θ), (3) where g s a nonlnear vector-valued mappng functon and θ s a vector of parameters. hen, an th entry s defned as follows: r ˆ = ( v, θ). (31) g here are several regresson algorthms 21 but only the regresson method based on the radal bass functon (RBF) s used n ths study for all methods. he reason s that the RBF method s relatvely fast and performs well. 12

13 Addtonal echnques All machne learnng algorthms may need the addtonal technques that help n parameter adjustment. he regressve PCA method used n ths study s a technque whch combnes the PCA and nonlnear regresson methods. 23 In general, the ways utlzed n both approaches to detect the underlyng dmensonalty of the data can be combned. For PCA, ths s an analyss of the resdual energy dependng on a number of PCs. Furthermore, for regresson methods ths s Automatc Relevance Determnaton (ARD). 21 he ARD method defnes the statstcal dependence between the PCs, and n the case of the dependency between the tested components and a target component the tested components are relevant to approxmate the target component. However, ths technque wll not be used n ths study. For the regressve PCA the number of real PCs wll be gven and a number of approxmated PCs wll be used as a free parameter. he pece-wse lnear Wener estmaton approach needs to determne the number of lnear components for usng a clusterng procedure. hs s done based on the model selecton of the mxed dstrbuton. 24 After that the Gaussan mxture model 21 wth a gven number of clusters s used to extract lnear components. Statstcal Propertes of Reflectance Spectra For statstcal analyss of the spectral reflectance of pantngs we use fve sets of color patches of ol or watercolor pant as follows: set A, 336 patches of pant (reflectance of pant); set B, 6 patches of pant (urner acryl gouache); set C, 6 patches of pant 13

14 (urner golden acrylcs); set D, 91 patches of pant (Kusakabe ol pant) and set E, 18 patches of pant (Kusakabe haban). All sets were extracted from the standard object color spectral database constructed by the Spectral Characterstc Database Constructon Workng Group. 25 hese sets have a spectral range of 4-7 nm and samples are evenly taken at 1 nm. he set A s used for tranng the algorthms and the sets B-E are used for predcton of the spectral reflectance. herefore, lnear and nonlnear prncpal component analyss was carred out only for the set A. Accordng to a prevous publcaton 3, fve PCs of lnear PCA are good enough for accurate spectral estmaton. Hence the spectral set A and ts frst fve PCs that have a resdual energy of.16% are analyzed and shown n Fg. 2 and Fg. 3, respectvely. 1 Set A.8 Reflectance Wavelength, nm Fg. 2. Reflectance spectra of the set A of pant patches. 14

15 .5 Frst PC.5 Second PC hrd PC Fourth PC Ffth PC Wavelength, nm Wavelength, nm Fg. 3. Frst fve prncpal components of the set A of pant patches. If regressve PCA s appled to utlze the fve real PCs and several approxmated PCs of the set A, the average RMSE value of the spectral approxmaton s reduced (Fg. 4). hs llustrates the fact that there s a way to mprove the degree of accuracy for representng spectra by ncorporatng the nonlnearty of the data. Experment Synthetc Data In ths secton, the nonlnear dataset s frst syntheszed and then all methods for spectral estmaton are tested wth a synthetc set. It s assumed that one channel response s used whle the data smulatng spectra s two-dmensonal. he purpose of the test s to show the feasblty of the method to work wth data whch has a nonlnear structure. 15

16 9.5 x RMSE Number of components Fg. 4. he average RMSE of spectral approxmaton for the set A usng regressve PCA. he frst fve components are gven by PCA and the components 6-1 are approxmated by regressve PCA. hus two data components are generated for the test. he frst component x1 s 4 unformly dstrbuted n the range and another one s x ( x.5. 2 = 1 ) Fnally, a zero-mean Gaussan nose wth the standard devaton.7 was added to the generated components. he estmaton result of the synthetc data s presented n Fg. 5. A vector F, a vector b 1, that s a frst PCA egenvector from B and the curve correspondng to an underlyng subspace are shown n Fg. 5. he orgnal (syntheszed) data and the estmates for each method are shown by gray dots n Fg. 5. Although the WE method s superor to the PCE based method, the PCE and WE methods gve poor estmates for the data. he MRE, RPCE and PLWE methods are relatvely good for estmaton. he RE method gves the best result from among these methods. 16

17 Orgnal data PCE WE MRE x F 1 F 1 F 1 F b b b b x 1 RPCE PLWE RE x F 1 F 1 F b b b x x x 1 Fg. 5. he estmaton results for the synthetc data and dfferent estmaton methods. Real Data An experment was conducted wth sets A-E descrbed above. he set A s used for tranng whle the other sets are used for predcton. he spectral transmttance characterstcs of the separaton flters used n a CCD camera are gven n Fg. 6. he spectral senstvty of a CCD area sensor (Phase One 372 (horzontal pxels) 26 (vertcal pxels), 14 bts) s presented n Fg. 7. he llumnaton source s D65. 17

18 ransmttance BPB 42 SP 9 BPB 55 BPB 5 BPB Wavelength, nm Fg. 6. he spectral transmttance characterstcs of the flters. 1.8 Senstvty Wavelength, nm Fg. 7. he spectral senstvty of the camera. he parameters used n the test are the followng: he fve PCs are exploted for PCE and RPCE. In addton, the RPCE approach uses the PCs approxmatng the real sxth, seventh, eghth and nnth PCs. For the PLWE method a mxture of Gaussan components s used for clusterng where the number of components s defned n a test based on the model selecton of the mxed dstrbuton. he MRE technque uses the terms begnnng wth the frst-order to the second-order ones. For the RE method, regresson s based on the radal bass functon usng the Gaussan functon. 2 neurons and 7 teratons are used n ths case. 18

19 A varatonal Bayesan model selecton method for the mxture dstrbuton 24 n the sensor space defnes the number of components for the PLWE method. For ths, the program s rerun ten tmes. he results are presented n able 1 where the frst row shows the test number and the second row shows the number of components determned by the algorthm. Fg. 8 llustrates the varatonal lkelhood bound over the model selecton of 336 pantng spectra (set A). Intally, the model has ten Gaussans. he vertcal lnes show the removal of the components from the model. Fnally, two components are selected. able 1. he number of components for pece-wse lnear Wener estmaton est number Number of components Lakelhood bound Mxture components Iteratons Fg. 8. he varatonal lkelhood bound over the model selecton of 336 pantng spectra (set A). 19

20 If the estmaton values of spectral reflectance are less than zero or greater than one then they are equalzed to zero or one, respectvely In able 2 and able 3, the average and maxmum RMSE values for each set are gven for the tradtonal methods and methods based on machne learnng algorthms. able 2. he average and maxmum (n parentheses) RMSE values for PCE, WE and MRE PCE WE MRE Set A.516 (.2458).155 (.1633).123 (.1159) Set B.836 (.3952).346 (.1712).324 (.1732) Set C.889 (.3469).466 (.2478).397 (.2158) Set D.917 (.483).43 (.234).352 (.275) Set E.917 (.3136).33 (.1416).281 (.1199) able 3. he average and maxmum (n parentheses) RMSE values for RPCE, PLWE and RE RPCE PLWE RE Set A.512 (.2447).142 (.1522).123 (.147) Set B.834 (.3928).343 (.1683).315 (.1731) Set C.887 (.3452).45 (.235).379 (.21) Set D.912 (.466).376 (.229).349 (.1992) Set E.91 (.3122).339 (.1185).275 (.162) 2

21 In able 4 and able 5, the average and maxmum CIE E 94 values for each set are gven for the tradtonal methods and methods based on machne learnng algorthms. able 4. he average and maxmum (n parentheses) CIE E 94 values for PCE, WE and MRE PCE WE MRE Set A.72 (13.65).17 (4.3).15 (1.68) Set B 2.96 (21.).58 (2.84).54 (2.13) Set C 2.36 (15.42).8 (4.8).59 (4.21) Set D 2.43 (19.24).71 (5.18).55 (3.37) Set E 1.32 (3.57).37 (2.34).31 (1.18) able 5. he average and maxmum (n parentheses) CIE E 94 values for RPCE, PLWE and RE RPCE PLWE RE Set A.81 (14.89).16 (3.46).17 (3.16) Set B 3.34 (23.15).67 (2.65).59 (2.65) Set C 2.51 (14.9) 1.33 (8.47).82 (3.47) Set D 2.71 (2.86).8623 (8.19).74 (2.92) Set E 1.89 (5.14).57 (2.).71 (2.79) In general, the results presented n able 2 - able 5 demonstrate that for the RMSE values the machne learnng methods gve slghtly better results than ther tradtonal opposte methods whle the tradtonal methods have smaller CIE E 94 values. he excepton s the RE method whch has better predcton n comparson wth the other methods for the maxmal error of the color dfference. 21

22 he methods are also tested usng computatonal tme. he CPU tme n seconds for the set A s presented n able 6. For the algorthms, the CPU tme s gven separately for tranng (upper row) and predcton (lower row). In able 6, zero values are gven for the CPU tme, whch s very small (ths corresponds to several matrx-vector multplcatons). Matlab 6.5, the Intel Pentum III Processor, 166 MHz and 248 MB of RAM are used n the test. he test shows that the tradtonal methods are faster than the machne learnng methods. However, the predcton tme for the machne learnng methods s relatvely short. able 6. he CPU tme n seconds PCE WE MRE RPCE PLWE RE o see whether any nonlnearty s presented n the estmated spectra we measure the average RMSE value after estmaton of spectral reflectance usng PCA and RPCA. he results are shown n able 7 for PCA wth the fve PCs (upper number) and for RPCA wth the fve real PCs and fve approxmated (from sx to ten) PCs (lower number). hen, the rato between these two RMSE values s determned and presented n able 8. From able 8, one can see that the RE and RPCE methods have rato values close to the orgnal data set. he MRE and PLWE methods gve results whch are farther from the orgnal data set. he PCE and WE rato values are the most dfferent from the orgnal data n comparson wth the other methods. 22

23 able 7. he average RMSE value after spectral estmaton for PCA wth the fve PCs (upper number) and for RPCA wth the fve real components and fve approxmated components (lower number). Set A PCE WE MRE RPCE PLWE RE able 8. he rato between the RMSE values for PCA and RPCA Set A PCE WE MRE RPCE PLWE RE From among the tradtonal methods the method based on MRE produces the best result. he method has small RMSE and CIE E 94 values n the tranng set and sets used for predcton. Whle the RMSE values for all machne learnng methods are slghtly better n comparson wth the tradtonal methods, the CIE E 94 values of the methods based on machne learnng except the RE method are hgher. he overall means of average color dfferences for the tradtonal methods are 1.95 (PCE),.52 (WE) and.42 (MRE) and for the learnng methods 2.25 (RPCE),.65 (PLWE) and.6 (RE). hus, the color dfferences usng the machne learnng methods are smaller than the dfferences between the tradtonal methods. he RE method ncorporates nonlnearty of data that s clearly seen from able 8. he generalzaton of the data gven by the RE method s very good n comparson wth the other methods. hs follows from predctng the maxmum CIE E 94 values. However, gven the processng and executon tmes the MRE method gves a better average and n two out of fve cases smaller maxmum color dfference errors than the RE method. Although 23

24 the tradtonal methods are less tme consumng than the machne learnng methods, the predcton tme for the learnng methods s short enough. In general, the tradtonal methods look more desrable than the learnng methods. hs s contrary to the ntal result shown n Fg. 5 where the learnng methods are superor to the tradtonal methods. hs can be explaned as follows. In ths study the sensor space (subspace) dmensonalty s defned by the fve gven flters. Although the subspace s not optmal (close to optmal) ts dmensonalty s rather hgh. Recently, t was shown that for reflectance spectra the dmensonalty of the nonlnear subspace s approxmately three. 26 hus, one can expect that for spectral magng systems havng the low dmensonal sensor space or fewer channels the learnng based methods are more effcent. We wll consder ths problem n a future study. Conclusons We have compared the methods for estmatng the spectral reflectance of art pantngs for the development of spectral color magng systems. hree tradtonal methods and three methods based on machne learnng for spectral reflectance estmaton of pant were utlzed. he tradtonal methods nclude two lnear methods the method based on PCA and the method based on Wener estmaton and one method usng multple regresson analyss. We ntroduced two novel machne learnng methods utlzng regressve PCA and pece-wse lnear Wener estmaton. hus, the machne learnng methods nclude two methods workng wth a global nonlnear data structure the method based on regressve PCA and the method based on regresson analyss and the method usng pece-wse lnear Wener estmaton. Smlarly to the submanfold method 2, the learnng methods used are between the lnear and Bayesan approaches 24

25 and the methods workng wth nonlnear data have a lmtaton. hey work only wth a data structure wth a one-to-one mappng. Fnally, we syntheszed a spectral color magng system mplementng the dfferent estmaton methods and demonstrated the possblty for accurately estmatng the reflectance spectra usng the presented technques. Acknowledgments he authors thank the Academy of Fnland for the fundng granted to ths study. References 1 Y. Myake, Evaluaton of Image Qualty Based on Human Vsual Characterstcs, n Proc. of the Frst Internatonal Workshop on Image Meda Qualty and ts Applcatons, Nagoya, Japan, pp (25). 2 Y. Myake, Y. Yokoyama, N. sumura, H. Hanesh, K. Myata, and J. Hayash, Development of Multband Color Imagng Systems for Recordng of Art Pantngs, n Color Imagng: Devce-Independent Color, Color Hardcopy, and Graphc Arts, Proc. of SPIE 3648, pp (1999). 3 H. Hanesh,. Hasegawa, A. Hoso, Y. Yokoyama, N. sumura and Y. Myake, System Desgn for Accurately Estmatng the Spectral Reflectance of Art Pantngs, Appled Optcs 39, (35) pp (2). 4 N. sumura, H. Hanesh, and Y. Myake, Estmaton of Spectral Reflectances from Mult-Band Images by Multple Regresson Analyss, Japanese Journal of Optcs, 27, (7) pp [n Japanese] (1998). 25

26 5 M. J. Vhrel and H. J. russel, Color Correcton Usng Prncpal Components, Color Res. App. 17, pp (1992). 6 M. J. Vrhel and H. J. russel, Flter Consderatons n Color Correcton, IEEE rans. Image Process. 3, pp (1994). 7 S. Goodall, P H. Lews, K. Martnez, P. A. S. Snclar, F. Gorgn, M. J. Adds, M. J. Bonface, C. Lahaner, and J. Stevenson, SCULPEUR: Multmeda Retreval for Museums, n Proc. of the Internatonal Conference Image and Vdeo Retreval CIVR 24, Dubln, Ireland, pp ( 24). 8 K. Martnez, J. Cuptt and D. Saunders, Hgh Resoluton Colormetrc Imagng of Pantngs, n Cameras, Scanners, and Image Acquston Systems, Proc. of SPIE 191, pp (1993). 9 J. E. Farrell, J. Cupptt, D. Saunders and B. A. Wandel, Estmatng Spectral Reflectances of Dgtal Images of Art, n Proc. of the Internatonal Symposum of Multspectral Imagng and Color Reproducton for Dgtal Archves, Chba, Japan, pp (1999). 1 J. Y. Hardeberg, H. Brettel and F. Schmtt, Spectral Characterzaton of Electronc Cameras, n Electronc Imagng: Processng, Prntng and Publshng n Color, Proc. of SPIE 349, pp (1998). 11 H. Maître, F. Schmtt, J.-P. Crettez, Y. Wu and J. Y. Hardeberg, Spectrophotometrc Image Analyss of Fne Art Pantngs, n Proc. of the Fourth Color Imagng Conference, Scottsdale, Arzona, pp (1996). 12 M. Hauta-Kasar, K. Myazava, S. oyooka, J. Parkknen and. Jaaskelanen, Spectral Vson System Based on Rewrtable Broad Band Color Flters, n Proc. of the 26

27 Internatonal Symposum of Multspectral Imagng and Color Reproducton for Dgtal Archves, Chba, Japan, pp (1999). 13 P. D. Burns, and R. S. Berns, Analyss of Multspectral Image Capture, n Proc. of the Fourth Color Imagng Conference, Scottsdale, Arzona, pp (1996). 14 F. H. Ima and R. S. Berns, Hgh-resoluton Mult-spectral Image Archves: a Hybrd Approach, n Proc. of the Fourth Color Imagng Conference, Scottsdale, Arzona, pp (1998). 15 L.. Maloney, Evaluaton of Lnear Models of Surface Spectral Reflectance wth Small Number of Parameters, J. Opt. Soc. Am. A 1, pp (1986). 16 J. Parkknen, J. Hallkanen and. Jaaskelanen, Characterstc Spectra of Munsell Color, J. Opt. Soc. Am. A 6, pp (1989). 17 M. J. Vrhel, R. Gershon, and L. S. Iwan, Measurement and Analyss of Object Reflectance Spectra, Color Res. App. 19, pp. 4-9 (1994). 18 A. Rbes, F. Schmtt and H. Brettel, Reconstructng Spectral Reflectances of Ol Pgments wth Neural Networks, n Proc. of the hrd Internatonal Conference on Multspectral Color Scence, Joensuu, Fnland, pp (21). 19 D. H. Branard and W.. Freeman, Bayesan Color Constancy, J. Opt. Soc. Am. A 14, pp (1997). 2 J. M. DCarlo and B. A. Wandell, Spectral Estmaton heory: Beyond Lnear but before Bayesan, J. Opt. Soc. Am. A 2, pp (23). 21 I.. Nabney, Netlab Algorthms for Pattern Recognton (Sprnger, 22) 22 Netlab oolbox, 27

28 23 V. Bochko and J. Parkknen, Prncpal Component Analyss Usng Approxmated Prncpal Components, Research Report 9, Department of Informaton echnology, Lappeenranta Unversty of echnology, pp. 1-7 (24). 24 A. Corduneanu, A. and C. M. Bshop, Varatonal Bayesan Model Selecton for Mxture Dstrbutons, n Proc. of the Eghth Internatonal Conference on Artfcal Intellgence and Statstcs, edted by. Rchardson and. Jaakkola, pp (Morgan Kaufmann, 21). 25 J. ajma, M. sukada, Y. Myake, H. Hanesh, N. sumura, M. Nakajma, Y. Azuma,. Iga, M. Inu, N. Ohta, N. Ojma and S. Sanada, Development and Standardzaton of a Spectral Characterstcs Data Base for Evaluatng Color Reproducton n Image Input Devces, n Electronc Imagng: Processng, Prntng, and Publshng n Color, Proc. of SPIE 349, pp ( 1998). 26 B. Funt, D. Kulpnsk, and V. Carde, Non-Lnear Embeddngs and the Underlyng Dmensonalty of Reflectance Spectra and Chromatcty Hstograms, n Proc. of the Nnth Color Imagng Conference: Color Scence, Systems and Applcatons, pp (21). 28

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

A Noise Analysis for Recovering Reflectances of the Objects Being Imaged

A Noise Analysis for Recovering Reflectances of the Objects Being Imaged A Nose Analyss for Recoverng Reflectances of the Objects Beng Imaged September 0 Mkya HIRONAGA A Nose Analyss for Recoverng Reflectances of the Objects Beng Imaged 画像入力による分光反射率復元に対するノイズの影響解析 September

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

Six-Band HDTV Camera System for Color Reproduction Based on Spectral Information

Six-Band HDTV Camera System for Color Reproduction Based on Spectral Information IS&T's 23 PICS Conference Sx-Band HDTV Camera System for Color Reproducton Based on Spectral Informaton Kenro Ohsawa )4), Hroyuk Fukuda ), Takeyuk Ajto 2),Yasuhro Komya 2), Hdeak Hanesh 3), Masahro Yamaguch

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Face Recognition University at Buffalo CSE666 Lecture Slides Resources:

Face Recognition University at Buffalo CSE666 Lecture Slides Resources: Face Recognton Unversty at Buffalo CSE666 Lecture Sldes Resources: http://www.face-rec.org/algorthms/ Overvew of face recognton algorthms Correlaton - Pxel based correspondence between two face mages Structural

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Edge Detection in Noisy Images Using the Support Vector Machines

Edge Detection in Noisy Images Using the Support Vector Machines Edge Detecton n Nosy Images Usng the Support Vector Machnes Hlaro Gómez-Moreno, Saturnno Maldonado-Bascón, Francsco López-Ferreras Sgnal Theory and Communcatons Department. Unversty of Alcalá Crta. Madrd-Barcelona

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

SVM-based Learning for Multiple Model Estimation

SVM-based Learning for Multiple Model Estimation SVM-based Learnng for Multple Model Estmaton Vladmr Cherkassky and Yunqan Ma Department of Electrcal and Computer Engneerng Unversty of Mnnesota Mnneapols, MN 55455 {cherkass,myq}@ece.umn.edu Abstract:

More information

Recovering spectral data from digital prints with an RGB camera using multi-exposure method

Recovering spectral data from digital prints with an RGB camera using multi-exposure method Recoverng spectral data from dgtal prnts wth an RGB camera usng mult-exposure method Mkko Nuutnen, Prkko Ottnen; Department of Meda Technology, Aalto Unversty School of Scence and Technology; Espoo, Fnland

More information

Lecture 4: Principal components

Lecture 4: Principal components /3/6 Lecture 4: Prncpal components 3..6 Multvarate lnear regresson MLR s optmal for the estmaton data...but poor for handlng collnear data Covarance matrx s not nvertble (large condton number) Robustness

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole

The Research of Ellipse Parameter Fitting Algorithm of Ultrasonic Imaging Logging in the Casing Hole Appled Mathematcs, 04, 5, 37-3 Publshed Onlne May 04 n ScRes. http://www.scrp.org/journal/am http://dx.do.org/0.436/am.04.584 The Research of Ellpse Parameter Fttng Algorthm of Ultrasonc Imagng Loggng

More information

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches Proceedngs of the Internatonal Conference on Cognton and Recognton Fuzzy Flterng Algorthms for Image Processng: Performance Evaluaton of Varous Approaches Rajoo Pandey and Umesh Ghanekar Department of

More information

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

A Robust Method for Estimating the Fundamental Matrix

A Robust Method for Estimating the Fundamental Matrix Proc. VIIth Dgtal Image Computng: Technques and Applcatons, Sun C., Talbot H., Ourseln S. and Adraansen T. (Eds.), 0- Dec. 003, Sydney A Robust Method for Estmatng the Fundamental Matrx C.L. Feng and Y.S.

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION

MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION MULTISPECTRAL IMAGES CLASSIFICATION BASED ON KLT AND ATR AUTOMATIC TARGET RECOGNITION Paulo Quntlano 1 & Antono Santa-Rosa 1 Federal Polce Department, Brasla, Brazl. E-mals: quntlano.pqs@dpf.gov.br and

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

Recognizing Faces. Outline

Recognizing Faces. Outline Recognzng Faces Drk Colbry Outlne Introducton and Motvaton Defnng a feature vector Prncpal Component Analyss Lnear Dscrmnate Analyss !"" #$""% http://www.nfotech.oulu.f/annual/2004 + &'()*) '+)* 2 ! &

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach

Skew Angle Estimation and Correction of Hand Written, Textual and Large areas of Non-Textual Document Images: A Novel Approach Angle Estmaton and Correcton of Hand Wrtten, Textual and Large areas of Non-Textual Document Images: A Novel Approach D.R.Ramesh Babu Pyush M Kumat Mahesh D Dhannawat PES Insttute of Technology Research

More information

NAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics

NAG Fortran Library Chapter Introduction. G10 Smoothing in Statistics Introducton G10 NAG Fortran Lbrary Chapter Introducton G10 Smoothng n Statstcs Contents 1 Scope of the Chapter... 2 2 Background to the Problems... 2 2.1 Smoothng Methods... 2 2.2 Smoothng Splnes and Regresson

More information

PCA Based Gait Segmentation

PCA Based Gait Segmentation Honggu L, Cupng Sh & Xngguo L PCA Based Gat Segmentaton PCA Based Gat Segmentaton Honggu L, Cupng Sh, and Xngguo L 2 Electronc Department, Physcs College, Yangzhou Unversty, 225002 Yangzhou, Chna 2 Department

More information

LECTURE : MANIFOLD LEARNING

LECTURE : MANIFOLD LEARNING LECTURE : MANIFOLD LEARNING Rta Osadchy Some sldes are due to L.Saul, V. C. Raykar, N. Verma Topcs PCA MDS IsoMap LLE EgenMaps Done! Dmensonalty Reducton Data representaton Inputs are real-valued vectors

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Hermite Splines in Lie Groups as Products of Geodesics

Hermite Splines in Lie Groups as Products of Geodesics Hermte Splnes n Le Groups as Products of Geodescs Ethan Eade Updated May 28, 2017 1 Introducton 1.1 Goal Ths document defnes a curve n the Le group G parametrzed by tme and by structural parameters n the

More information

Hybridization of Expectation-Maximization and K-Means Algorithms for Better Clustering Performance

Hybridization of Expectation-Maximization and K-Means Algorithms for Better Clustering Performance BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 2 Sofa 2016 Prnt ISSN: 1311-9702; Onlne ISSN: 1314-4081 DOI: 10.1515/cat-2016-0017 Hybrdzaton of Expectaton-Maxmzaton

More information

Classification / Regression Support Vector Machines

Classification / Regression Support Vector Machines Classfcaton / Regresson Support Vector Machnes Jeff Howbert Introducton to Machne Learnng Wnter 04 Topcs SVM classfers for lnearly separable classes SVM classfers for non-lnearly separable classes SVM

More information

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS

EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS P.G. Demdov Yaroslavl State Unversty Anatoly Ntn, Vladmr Khryashchev, Olga Stepanova, Igor Kostern EYE CENTER LOCALIZATION ON A FACIAL IMAGE BASED ON MULTI-BLOCK LOCAL BINARY PATTERNS Yaroslavl, 2015 Eye

More information

A Background Subtraction for a Vision-based User Interface *

A Background Subtraction for a Vision-based User Interface * A Background Subtracton for a Vson-based User Interface * Dongpyo Hong and Woontack Woo KJIST U-VR Lab. {dhon wwoo}@kjst.ac.kr Abstract In ths paper, we propose a robust and effcent background subtracton

More information

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines

A Modified Median Filter for the Removal of Impulse Noise Based on the Support Vector Machines A Modfed Medan Flter for the Removal of Impulse Nose Based on the Support Vector Machnes H. GOMEZ-MORENO, S. MALDONADO-BASCON, F. LOPEZ-FERRERAS, M. UTRILLA- MANSO AND P. GIL-JIMENEZ Departamento de Teoría

More information

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval

Fuzzy C-Means Initialized by Fixed Threshold Clustering for Improving Image Retrieval Fuzzy -Means Intalzed by Fxed Threshold lusterng for Improvng Image Retreval NAWARA HANSIRI, SIRIPORN SUPRATID,HOM KIMPAN 3 Faculty of Informaton Technology Rangst Unversty Muang-Ake, Paholyotn Road, Patumtan,

More information

Learning-Based Top-N Selection Query Evaluation over Relational Databases

Learning-Based Top-N Selection Query Evaluation over Relational Databases Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **

More information

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal

More information

Laplacian Eigenmap for Image Retrieval

Laplacian Eigenmap for Image Retrieval Laplacan Egenmap for Image Retreval Xaofe He Partha Nyog Department of Computer Scence The Unversty of Chcago, 1100 E 58 th Street, Chcago, IL 60637 ABSTRACT Dmensonalty reducton has been receved much

More information

Lecture 5: Multilayer Perceptrons

Lecture 5: Multilayer Perceptrons Lecture 5: Multlayer Perceptrons Roger Grosse 1 Introducton So far, we ve only talked about lnear models: lnear regresson and lnear bnary classfers. We noted that there are functons that can t be represented

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Learning an Image Manifold for Retrieval

Learning an Image Manifold for Retrieval Learnng an Image Manfold for Retreval Xaofe He*, We-Yng Ma, and Hong-Jang Zhang Mcrosoft Research Asa Bejng, Chna, 100080 {wyma,hjzhang}@mcrosoft.com *Department of Computer Scence, The Unversty of Chcago

More information

Principal Component Inversion

Principal Component Inversion Prncpal Component Inverson Dr. A. Neumann, H. Krawczyk German Aerospace Centre DLR Remote Sensng Technology Insttute Marne Remote Sensng Prncpal Components - Propertes The Lnear Inverson Algorthm Optmsaton

More information

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1

A New Feature of Uniformity of Image Texture Directions Coinciding with the Human Eyes Perception 1 A New Feature of Unformty of Image Texture Drectons Concdng wth the Human Eyes Percepton Xng-Jan He, De-Shuang Huang, Yue Zhang, Tat-Mng Lo 2, and Mchael R. Lyu 3 Intellgent Computng Lab, Insttute of Intellgent

More information

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET Jae-young Lee, Shahram Payandeh, and Ljljana Trajovć School of Engneerng Scence Smon Fraser Unversty 8888 Unversty

More information

CS 534: Computer Vision Model Fitting

CS 534: Computer Vision Model Fitting CS 534: Computer Vson Model Fttng Sprng 004 Ahmed Elgammal Dept of Computer Scence CS 534 Model Fttng - 1 Outlnes Model fttng s mportant Least-squares fttng Maxmum lkelhood estmaton MAP estmaton Robust

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

More information

Bootstrapping Color Constancy

Bootstrapping Color Constancy Bootstrappng Color Constancy Bran Funt and Vlad C. Carde * Smon Fraser Unversty Vancouver, Canada ABSTRACT Bootstrappng provdes a novel approach to tranng a neural network to estmate the chromatcty of

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

Smoothing Spline ANOVA for variable screening

Smoothing Spline ANOVA for variable screening Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory

More information

Face Recognition Based on SVM and 2DPCA

Face Recognition Based on SVM and 2DPCA Vol. 4, o. 3, September, 2011 Face Recognton Based on SVM and 2DPCA Tha Hoang Le, Len Bu Faculty of Informaton Technology, HCMC Unversty of Scence Faculty of Informaton Scences and Engneerng, Unversty

More information

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity

Corner-Based Image Alignment using Pyramid Structure with Gradient Vector Similarity Journal of Sgnal and Informaton Processng, 013, 4, 114-119 do:10.436/jsp.013.43b00 Publshed Onlne August 013 (http://www.scrp.org/journal/jsp) Corner-Based Image Algnment usng Pyramd Structure wth Gradent

More information

Adaptive Transfer Learning

Adaptive Transfer Learning Adaptve Transfer Learnng Bn Cao, Snno Jaln Pan, Yu Zhang, Dt-Yan Yeung, Qang Yang Hong Kong Unversty of Scence and Technology Clear Water Bay, Kowloon, Hong Kong {caobn,snnopan,zhangyu,dyyeung,qyang}@cse.ust.hk

More information

Human Face Recognition Using Generalized. Kernel Fisher Discriminant

Human Face Recognition Using Generalized. Kernel Fisher Discriminant Human Face Recognton Usng Generalzed Kernel Fsher Dscrmnant ng-yu Sun,2 De-Shuang Huang Ln Guo. Insttute of Intellgent Machnes, Chnese Academy of Scences, P.O.ox 30, Hefe, Anhu, Chna. 2. Department of

More information

THE PULL-PUSH ALGORITHM REVISITED

THE PULL-PUSH ALGORITHM REVISITED THE PULL-PUSH ALGORITHM REVISITED Improvements, Computaton of Pont Denstes, and GPU Implementaton Martn Kraus Computer Graphcs & Vsualzaton Group, Technsche Unverstät München, Boltzmannstraße 3, 85748

More information

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010

Simulation: Solving Dynamic Models ABE 5646 Week 11 Chapter 2, Spring 2010 Smulaton: Solvng Dynamc Models ABE 5646 Week Chapter 2, Sprng 200 Week Descrpton Readng Materal Mar 5- Mar 9 Evaluatng [Crop] Models Comparng a model wth data - Graphcal, errors - Measures of agreement

More information

Fuzzy Logic Based RS Image Classification Using Maximum Likelihood and Mahalanobis Distance Classifiers

Fuzzy Logic Based RS Image Classification Using Maximum Likelihood and Mahalanobis Distance Classifiers Research Artcle Internatonal Journal of Current Engneerng and Technology ISSN 77-46 3 INPRESSCO. All Rghts Reserved. Avalable at http://npressco.com/category/jcet Fuzzy Logc Based RS Image Usng Maxmum

More information

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET

APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET APPLICATION OF PREDICTION-BASED PARTICLE FILTERS FOR TELEOPERATIONS OVER THE INTERNET Jae-young Lee, Shahram Payandeh, and Ljljana Trajovć School of Engneerng Scence Smon Fraser Unversty 8888 Unversty

More information

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc.

An Indian Journal FULL PAPER ABSTRACT KEYWORDS. Trade Science Inc. [Type text] [Type text] [Type text] ISSN : 97-735 Volume Issue 9 BoTechnology An Indan Journal FULL PAPER BTAIJ, (9), [333-3] Matlab mult-dmensonal model-based - 3 Chnese football assocaton super league

More information

Wavefront Reconstructor

Wavefront Reconstructor A Dstrbuted Smplex B-Splne Based Wavefront Reconstructor Coen de Vsser and Mchel Verhaegen 14-12-201212 2012 Delft Unversty of Technology Contents Introducton Wavefront reconstructon usng Smplex B-Splnes

More information

Dependence of the Color Rendering Index on the Luminance of Light Sources and Munsell Samples

Dependence of the Color Rendering Index on the Luminance of Light Sources and Munsell Samples Australan Journal of Basc and Appled Scences, 4(10): 4609-4613, 2010 ISSN 1991-8178 Dependence of the Color Renderng Index on the Lumnance of Lght Sources and Munsell Samples 1 A. EL-Bally (Physcs Department),

More information

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images

Using Fuzzy Logic to Enhance the Large Size Remote Sensing Images Internatonal Journal of Informaton and Electroncs Engneerng Vol. 5 No. 6 November 015 Usng Fuzzy Logc to Enhance the Large Sze Remote Sensng Images Trung Nguyen Tu Huy Ngo Hoang and Thoa Vu Van Abstract

More information

A Workflow for Spatial Uncertainty Quantification using Distances and Kernels

A Workflow for Spatial Uncertainty Quantification using Distances and Kernels A Workflow for Spatal Uncertanty Quantfcaton usng Dstances and Kernels Célne Schedt and Jef Caers Stanford Center for Reservor Forecastng Stanford Unversty Abstract Assessng uncertanty n reservor performance

More information

Three supervised learning methods on pen digits character recognition dataset

Three supervised learning methods on pen digits character recognition dataset Three supervsed learnng methods on pen dgts character recognton dataset Chrs Flezach Department of Computer Scence and Engneerng Unversty of Calforna, San Dego San Dego, CA 92093 cflezac@cs.ucsd.edu Satoru

More information

Recommended Items Rating Prediction based on RBF Neural Network Optimized by PSO Algorithm

Recommended Items Rating Prediction based on RBF Neural Network Optimized by PSO Algorithm Recommended Items Ratng Predcton based on RBF Neural Network Optmzed by PSO Algorthm Chengfang Tan, Cayn Wang, Yuln L and Xx Q Abstract In order to mtgate the data sparsty and cold-start problems of recommendaton

More information

Learning a Class-Specific Dictionary for Facial Expression Recognition

Learning a Class-Specific Dictionary for Facial Expression Recognition BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 16, No 4 Sofa 016 Prnt ISSN: 1311-970; Onlne ISSN: 1314-4081 DOI: 10.1515/cat-016-0067 Learnng a Class-Specfc Dctonary for

More information

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes

R s s f. m y s. SPH3UW Unit 7.3 Spherical Concave Mirrors Page 1 of 12. Notes SPH3UW Unt 7.3 Sphercal Concave Mrrors Page 1 of 1 Notes Physcs Tool box Concave Mrror If the reflectng surface takes place on the nner surface of the sphercal shape so that the centre of the mrror bulges

More information

Relevance Assignment and Fusion of Multiple Learning Methods Applied to Remote Sensing Image Analysis

Relevance Assignment and Fusion of Multiple Learning Methods Applied to Remote Sensing Image Analysis Assgnment and Fuson of Multple Learnng Methods Appled to Remote Sensng Image Analyss Peter Bajcsy, We-Wen Feng and Praveen Kumar Natonal Center for Supercomputng Applcaton (NCSA), Unversty of Illnos at

More information

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems

A Unified Framework for Semantics and Feature Based Relevance Feedback in Image Retrieval Systems A Unfed Framework for Semantcs and Feature Based Relevance Feedback n Image Retreval Systems Ye Lu *, Chunhu Hu 2, Xngquan Zhu 3*, HongJang Zhang 2, Qang Yang * School of Computng Scence Smon Fraser Unversty

More information

Lecture 13: High-dimensional Images

Lecture 13: High-dimensional Images Lec : Hgh-dmensonal Images Grayscale Images Lecture : Hgh-dmensonal Images Math 90 Prof. Todd Wttman The Ctadel A grayscale mage s an nteger-valued D matrx. An 8-bt mage takes on values between 0 and 55.

More information

An Ensemble Learning algorithm for Blind Signal Separation Problem

An Ensemble Learning algorithm for Blind Signal Separation Problem An Ensemble Learnng algorthm for Blnd Sgnal Separaton Problem Yan L 1 and Peng Wen 1 Department of Mathematcs and Computng, Faculty of Engneerng and Surveyng The Unversty of Southern Queensland, Queensland,

More information

A Robust LS-SVM Regression

A Robust LS-SVM Regression PROCEEDIGS OF WORLD ACADEMY OF SCIECE, EGIEERIG AD ECHOLOGY VOLUME 7 AUGUS 5 ISS 37- A Robust LS-SVM Regresson József Valyon, and Gábor Horváth Abstract In comparson to the orgnal SVM, whch nvolves a quadratc

More information

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1

Outline. Discriminative classifiers for image recognition. Where in the World? A nearest neighbor recognition example 4/14/2011. CS 376 Lecture 22 1 4/14/011 Outlne Dscrmnatve classfers for mage recognton Wednesday, Aprl 13 Krsten Grauman UT-Austn Last tme: wndow-based generc obect detecton basc ppelne face detecton wth boostng as case study Today:

More information

An Improved Image Segmentation Algorithm Based on the Otsu Method

An Improved Image Segmentation Algorithm Based on the Otsu Method 3th ACIS Internatonal Conference on Software Engneerng, Artfcal Intellgence, Networkng arallel/dstrbuted Computng An Improved Image Segmentaton Algorthm Based on the Otsu Method Mengxng Huang, enjao Yu,

More information

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning

Outline. Type of Machine Learning. Examples of Application. Unsupervised Learning Outlne Artfcal Intellgence and ts applcatons Lecture 8 Unsupervsed Learnng Professor Danel Yeung danyeung@eee.org Dr. Patrck Chan patrckchan@eee.org South Chna Unversty of Technology, Chna Introducton

More information

Snakes-based approach for extraction of building roof contours from digital aerial images

Snakes-based approach for extraction of building roof contours from digital aerial images Snakes-based approach for extracton of buldng roof contours from dgtal aeral mages Alur P. Dal Poz and Antono J. Fazan São Paulo State Unversty Dept. of Cartography, R. Roberto Smonsen 305 19060-900 Presdente

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

More information

Categories and Subject Descriptors B.7.2 [Integrated Circuits]: Design Aids Verification. General Terms Algorithms

Categories and Subject Descriptors B.7.2 [Integrated Circuits]: Design Aids Verification. General Terms Algorithms 3. Fndng Determnstc Soluton from Underdetermned Equaton: Large-Scale Performance Modelng by Least Angle Regresson Xn L ECE Department, Carnege Mellon Unversty Forbs Avenue, Pttsburgh, PA 3 xnl@ece.cmu.edu

More information

Histogram-Enhanced Principal Component Analysis for Face Recognition

Histogram-Enhanced Principal Component Analysis for Face Recognition Hstogram-Enhanced Prncpal Component Analyss for Face ecognton Ana-ara Sevcenco and Wu-Sheng Lu Dept. of Electrcal and Computer Engneerng Unversty of Vctora sevcenco@engr.uvc.ca, wslu@ece.uvc.ca Abstract

More information

The Study of Remote Sensing Image Classification Based on Support Vector Machine

The Study of Remote Sensing Image Classification Based on Support Vector Machine Sensors & Transducers 03 by IFSA http://www.sensorsportal.com The Study of Remote Sensng Image Classfcaton Based on Support Vector Machne, ZHANG Jan-Hua Key Research Insttute of Yellow Rver Cvlzaton and

More information

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

More information

What are the camera parameters? Where are the light sources? What is the mapping from radiance to pixel color? Want to solve for 3D geometry

What are the camera parameters? Where are the light sources? What is the mapping from radiance to pixel color? Want to solve for 3D geometry Today: Calbraton What are the camera parameters? Where are the lght sources? What s the mappng from radance to pel color? Why Calbrate? Want to solve for D geometry Alternatve approach Solve for D shape

More information

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data

Type-2 Fuzzy Non-uniform Rational B-spline Model with Type-2 Fuzzy Data Malaysan Journal of Mathematcal Scences 11(S) Aprl : 35 46 (2017) Specal Issue: The 2nd Internatonal Conference and Workshop on Mathematcal Analyss (ICWOMA 2016) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES

More information

Backpropagation: In Search of Performance Parameters

Backpropagation: In Search of Performance Parameters Bacpropagaton: In Search of Performance Parameters ANIL KUMAR ENUMULAPALLY, LINGGUO BU, and KHOSROW KAIKHAH, Ph.D. Computer Scence Department Texas State Unversty-San Marcos San Marcos, TX-78666 USA ae049@txstate.edu,

More information

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law)

Machine Learning. Support Vector Machines. (contains material adapted from talks by Constantin F. Aliferis & Ioannis Tsamardinos, and Martin Law) Machne Learnng Support Vector Machnes (contans materal adapted from talks by Constantn F. Alfers & Ioanns Tsamardnos, and Martn Law) Bryan Pardo, Machne Learnng: EECS 349 Fall 2014 Support Vector Machnes

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

COMPLEX WAVELET TRANSFORM-BASED COLOR INDEXING FOR CONTENT-BASED IMAGE RETRIEVAL

COMPLEX WAVELET TRANSFORM-BASED COLOR INDEXING FOR CONTENT-BASED IMAGE RETRIEVAL COMPLEX WAVELET TRANSFORM-BASED COLOR INDEXING FOR CONTENT-BASED IMAGE RETRIEVAL Nader Safavan and Shohreh Kasae Department of Computer Engneerng Sharf Unversty of Technology Tehran, Iran skasae@sharf.edu

More information

TN348: Openlab Module - Colocalization

TN348: Openlab Module - Colocalization TN348: Openlab Module - Colocalzaton Topc The Colocalzaton module provdes the faclty to vsualze and quantfy colocalzaton between pars of mages. The Colocalzaton wndow contans a prevew of the two mages

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

More information

Competitive Sparse Representation Classification for Face Recognition

Competitive Sparse Representation Classification for Face Recognition Vol. 6, No. 8, 05 Compettve Sparse Representaton Classfcaton for Face Recognton Yng Lu Chongqng Key Laboratory of Computatonal Intellgence Chongqng Unversty of Posts and elecommuncatons Chongqng, Chna

More information

Integrated Expression-Invariant Face Recognition with Constrained Optical Flow

Integrated Expression-Invariant Face Recognition with Constrained Optical Flow Integrated Expresson-Invarant Face Recognton wth Constraned Optcal Flow Chao-Kue Hseh, Shang-Hong La 2, and Yung-Chang Chen Department of Electrcal Engneerng, Natonal Tsng Hua Unversty, Tawan 2 Department

More information

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

More information

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros.

Fitting & Matching. Lecture 4 Prof. Bregler. Slides from: S. Lazebnik, S. Seitz, M. Pollefeys, A. Effros. Fttng & Matchng Lecture 4 Prof. Bregler Sldes from: S. Lazebnk, S. Setz, M. Pollefeys, A. Effros. How do we buld panorama? We need to match (algn) mages Matchng wth Features Detect feature ponts n both

More information